Artificial intelligence reveals past climate extremes by reconstructing historical records

Abstract

The understanding of recent climate extremes and the characterization of climate risk require examining these extremes within a historical context. However, the existing datasets of observed extremes generally exhibit spatial gaps and inaccuracies due to inadequate spatial extrapolation. This problem arises from traditional statistical methods used to account for the lack of measurements, particularly prevalent before the mid-20th century. In this work, we use artificial intelligence to reconstruct observations of European climate extremes (warm and cold days and nights) by leveraging Earth system model data from CMIP6 through transfer learning. Our method surpasses conventional statistical techniques and diffusion models, showcasing its ability to reconstruct past extreme events and reveal spatial trends across an extensive time span (1901-2018) that is not covered by most reanalysis datasets. Providing our dataset to the climate community will improve the characterization of climate extremes, resulting in better risk management and policies.

Overview of mean and extreme precipitation climate changes across the Dinaric Alps in the latest EURO-CORDEX ensemble

Abstract

The study provides a detailed analysis of the climate change projections in mean and extreme precipitation across the Dinaric Alps and the Adriatic coastal area until the end of the 21st century. It uses simulations from the extensive EURO-CORDEX regional climate model ensemble to study precipitation changes considering three greenhouse gas concentration scenarios. Additionally, the performance and systematic errors of historical simulations are evaluated. The ensemble demonstrates good skill in representing spatial variability and seasonal variations of mean and extreme precipitation. However, biases are evident and substantial across the Dinaric Alps, predominantly wet in winter and autumn, with the exceptions of dry biases in summer. The ensemble overestimates the frequency of heavy and extreme events. Regardless of these inconsistencies, projections clearly suggest a change in precipitation character with an overall intensification and a decrease in wet-day frequency, resulting in a mean precipitation winter increase over northern lowlands, summer decrease across southern parts, and spring and autumn zero-change zone across the Dinaric Alps. Extreme precipitation events are expected to intensify and become more frequent during winter and autumn with robust signals over the lowlands. During summer, the ensemble shows substantial uncertainties, but an intensification nonetheless within a smaller number of extreme events. Overall, the study identifies more consistency in the direction of change than magnitude in individual simulations, with the strongest consensus on precipitation intensification. Limitations include low station density in the observational dataset and an incomplete ensemble size, however, findings align with previous research and observed trends.

Near-term prediction of surface temperature extremes over India in the CMIP6-DCPP models

Abstract

Decadal climate prediction project (DCPP) hindcasts/predictions from Coupled Model Intercomparison Project phase-6 (CMIP6) provide vital information on the near-future climate up to a decade. Coarse-resolution climate models however fail to accurately represent the regional climatic features thereby limiting the prediction skill. In this study, DCPP models maximum and minimum surface temperatures (Tmax and Tmin) hindcasts are downscaled and bias-corrected (DBC) over the Indian region to examine the characteristics of heat/cold waves for 1–10 lead years. It is found that the DBC Tmax (Tmin) captures the heat (cold) waves intensity, frequency, and spatial distribution over India quite effectively. After DBC, representation of extreme thresholds of Tmax (Tmin) over India has improved by 82(90)%. Further, the areal extent of extremely high (low) temperatures associated with the large and small area heat (cold) waves are well characterized after DBC up to 10-year lead. Importantly, DBC product showed superior skills in capturing the regional temperature peaks associated with large and small area heatwave/cold wave days and is limited before DBC. After DBC, the temperature extremes (both warm and cold) display enhanced intensity with increased (decreased) mean Tmax (Tmin) by ~ 0.6 °C (0.3 °C) in the recent decade (2017–2026) compared to the previous decade (2007–2016) during April to June (November to February). The present analysis of DCPP models opens the door for the near future or decadal prediction of temperature extremes by demonstrating the increased prediction ability across India for Tmax and Tmin following DBC. The study has important implications for decision-making and may help different stakeholders, policymakers, and disaster managers.

Advancements in Environmental Data Analysis for Climate-Resilient Agriculture Using Remote Sensing and Deep Learning

Abstract

Weather significantly influences agricultural productivity. Plant biotic and abiotic stressors are primarily induced by climate change, resulting in a detrimental effect on worldwide agricultural productivity. These two components are closely interconnected. This paper presents an innovative approach in smart agriculture for climate change combining remote sensing and a deep learning algorithm. The input is obtained as a multispectral environmental image and then subjected to noise reduction and normalisation processing. The image has been retrieved using a primary convolutional component with a stacked encoder model. The retrieved features are identified using ResNet graph reinforcement neural networks. The resulting classification output displays environmental imagery depicting climate variations. The agriculture sector has been analysed based on the classified climate analysis. The experimental results have been conducted on diverse farm datasets related to climate change, evaluating the detection accuracy, recall, mean average precision, normalized correlation, and F-1 score. The proposed method achieved a detection accuracy of 98%, a normalised correlation of 95%, a mean average precision of 92%, a recall of 97%, and an F-MEASURE of 94%. Machine learning can assist in monitoring and predicting the impact of climate change on food security, as indicated by the findings.

Multivariate stochastic generation of meteorological data for building simulation through interdependent meteorological processes

Abstract

In recent years, the uncertainty of weather conditions and the impact of future climate change on building energy assessment has received increasing attention. As an important part of these studies, several types of methods for generating stochastic meteorological data have also been developed. Since solar radiation drops to zero at night, unlike the continuous 24-hour data for elements such as temperature and humidity, this has posed challenges for previous research to fully account for the simultaneity among multiple elements. Therefore, this study proposes a framework for meteorological data generation: First, perform multivariate time series modeling of meteorological data of air temperature, solar radiation and absolute humidity at 12:00 of each day of a typical year based on the S-vine copula method and simulating daily series data at 12:00 for 365 days. Then, based on the probability of change of each element evaluated from the historical meteorological observation data, the daily series data at 12:00 were expanded to 24 h, after which the yearly stochastic weather data were obtained. The analysis of 30 years of stochastic data generated by this method, compared with the original data, reveals that air temperature and solar radiation closely match the original distribution characteristics, except for a minor deviation in the absolute humidity’s kurtosis. Furthermore, the comparison of thermal load distributions for office buildings shows that the original data curve falls within the range of the generated data. This suggests that the generated data includes more information about uncertainty but still keeps the original data’s characteristics.

Integrated understanding of climate change and disaster risk for building resilience of cultural heritage sites

Abstract

Heritage assets are vulnerable to climate change and disaster risks. However, existing literature has long been separating climate change from disaster risks, which were mainly considered as natural disasters. Recently, the framework of integrated understanding of climate change and disaster risk reduction in international policies started to be discussed in sustainable development discussion, while mentioning opportunities to build resilience of cultural heritage sites (United Nations Office for Disaster Risk Reduction 2020). But this framework is yet to be implemented and detailed in the context of heritage sites. Therefore, the aim of this paper is to analyze how the integrated understanding of climate change and disaster risk reduction policies can contribute to building climate resilience of cultural heritage sites by reviewing the key themes emerging from the literature. The question this paper answers are how can the integrated understanding of climate change and disaster risks reduction tackle barriers to the resilience of heritage sites? And what can be done to fill the gaps identified in the literature? To understand it, four elements from the literature are analyzed, including methodological contributions, temporalities, challenges and gaps, and opportunities. The findings of this review help in understanding the gap and interplay between science and policy in decision-making processes. We conclude by discussing the ways forward for the applicability of the framework in building resilience of cultural heritage sites.

Fixing active sand dune by native grasses in the desert of Northwest China

Abstract

Background

Desertification is the most severe environmental problem in arid and semi-arid regions and has caused great economic loss every year. However, artificial sand fixation barriers function on sand fixation for only 10–20 years. Searching for a native species with long-term sand fixation effect and strong environmental adaptive capacity, and low water consumption is needed. In this study, we investigated the environmental adaption and sand fixation effect of a grass from Poaceae family (Psammochloa villosa) that is indigenous to the desert of Northwest China.

Results

The results showed that P. villosa has a streamlined leaf form, strong mechanical strength, and flexibility to adapt to wind. Leaf curling of P. villosa under drought decreased water loss rate through decreased evaporation area to adapt to drought. Significant negative relationship between adventitious root length and horizontal root burial depth indicates that adventitious roots help P. villosa absorb water and nutrients from soil under shallow sand burial condition, which enables P. villosa to adapt to different sand burial conditions. P. villosa fixed sand dunes through the distribution of the population at the top of the dune and the vertical relationship between the direction of windblown sand and the direction of growth of P. villosa, which stopped the expansion of the dune.

Conclusions

Growth characteristics of wind and drought tolerant leaf traits and adventitious roots under sand burial indicate that P. villosa is well adapted to dry sandy desert conditions and burial by sand. The distribution of the P. villosa population on the sand dune is a “brake” on its expansion. These findings provide new insight for active sand dune fixation and desertification control using native grass in the desertified regions.

Attributing human mortality from fire PM2.5 to climate change

Abstract

Climate change intensifies fire smoke, emitting hazardous air pollutants that impact human health. However, the global influence of climate change on fire-induced health impacts remains unquantified. Here we used three well-tested fire–vegetation models in combination with a chemical transport model and health risk assessment framework to attribute global human mortality from fire fine particulate matter (PM2.5) emissions to climate change. Of the 46,401 (1960s) to 98,748 (2010s) annual fire PM2.5 mortalities, 669 (1.2%, 1960s) to 12,566 (12.8%, 2010s) were attributed to climate change. The most substantial influence of climate change on fire mortality occurred in South America, Australia and Europe, coinciding with decreased relative humidity and in boreal forests with increased air temperature. Increasing relative humidity lowered fire mortality in other regions, such as South Asia. Our study highlights the role of climate change in fire mortality, aiding public health authorities in spatial targeting adaptation measures for sensitive fire-prone areas.

Experimental Evaluation of Remote Sensing–Based Climate Change Prediction Using Enhanced Deep Learning Strategy

Abstract

Climate change is one of the most pressing global challenges of our time, with far-reaching impacts on ecosystems, economies, and human societies. Accurate prediction of climate change patterns is crucial for developing effective mitigation and adaptation strategies. Remote sensing data, with its ability to provide comprehensive and continuous observations of the Earth’s surface, plays a vital role in monitoring and predicting these changes. However, the complexity and high dimensionality of remote sensing data present significant challenges for traditional predictive models. In this study, we present an Enhanced Deep Learning Strategy for climate change prediction using remote sensing data, integrating a Cascaded Inception-LGBM model. The proposed model combines the feature extraction capabilities of the Inception module with the predictive power of the Light Gradient Boosting Machine (LGBM). The methodology was evaluated on various climate variables, including temperature, precipitation, and CO2 levels, achieving an accuracy of 97.22%. Comparative analysis with state-of-the-art models demonstrated the superior performance of our approach, particularly in terms of RMSE, MAE, and R2 metrics. Robustness tests further confirmed the model’s generalization capabilities under different data conditions. This study underscores the potential of advanced deep learning techniques in enhancing climate change prediction accuracy and offers insights into the key drivers of climate variability.